Adoption trebled in under three years. The number of tools each business actually uses went from 1.4 to 1.6.
Take the headline first, because it is real. The share of British businesses with ten or more employees using at least one AI technology has gone from around 12 per cent in late 2023 to around 35 per cent by June this year. Adoption has roughly trebled in under three years. On any ordinary reading of a technology curve, that is fast.
Now take the second number, which was published on the same day in the same bulletin and travelled nowhere. The average number of AI technologies used by an adopting business rose over the same period from about 1.4 to about 1.6.
Three times as many firms, using barely more each. Only around 10 per cent of adopters describe their use as extensive.
The Office for National Statistics draws the obvious conclusion itself, and its wording is careful: this would imply relatively limited transformative impacts to date. That is a national statistical office saying, in the flat register those offices use, that the thing everyone is arguing about has not happened yet.
For a business, breadth and depth are different purchases. Adding a first tool is cheap, reversible and often free at the point of use. Somebody signs up, it helps with a task, the firm now counts as an adopter. Going deep is a different order of commitment: it means changing how work moves through the place, retraining people, rebuilding a process that currently functions. The first is a decision one person makes in an afternoon. The second is a decision a business makes over a year.
Britain has made the first decision three times over and the second hardly at all.
There is a size pattern inside it that runs the opposite way to expectation. Among businesses with nine employees or fewer, 17 per cent of adopters report extensive use. Among those with 250 or more, the figure is 9 per cent. The smallest firms that take this up take it up harder, which is the reverse of how technology usually spreads through an economy, and it has had almost no attention.
What follows for a reader is a question about timing rather than direction. If adoption is broad and shallow, then the effects being argued about are mostly still ahead. The people telling you the transformation has already happened are describing 10 per cent of 35 per cent of firms with ten or more staff, in an economy with six sectors excluded from the count.
That is not nothing. It is also not the thing it is being sold as.
The gain cannot have gone anywhere yet at scale, because the depth of use required to produce it does not exist in the data. Broad and shallow adoption is a statement about when, not whether.
- ONS, Artificial intelligence in UK businesses 2023 to 2026, published 20 July 2026. Section 3 for adoption rates, intensity of use, and the size-band breakdown of extensive use.
- Three figures in an earlier internal pass on this theme were second hand and wrong: adoption was recorded as 26 per cent from a March wave rather than around 35 per cent at Wave 159, and a 1 per cent headcount increase figure was misattributed to the smallest firms when it belongs to the subgroup using AI to improve operations. Corrected against the bulletin before drafting.